[racket-dev] Single-flonum-ness not preserved in racket/math functions

At Tue, 05 Jun 2012 22:05:26 -0600,
Neil Toronto wrote:
> Thanks! I've adapted it and will commit the changes with the ones to 
> racket/math.

Great, thanks!

> I've managed to make it always find a counterexample. But there are only 
> a few single-flonum -> flonum issues. Most of them are because I've 
> changed the random-flonum-generating code to sometimes generate *very* 
> small random numbers.

Good idea.

> Here's the counterexample I found earlier today 
> that motivated that change to the random testing:
> 
> #lang typed/racket
> 
> (require/typed
>   unstable/flonum
>   [+min.0  Positive-Float])  ; the smallest positive flonum
> 
>  > +min.0
> - : Float [generalized from Positive-Float]
> 4.9406564584125e-324
>  > (sqr +min.0)
> - : Float [generalized from Positive-Float]
> 0.0
> 
> 
> I'm not comfortable messing with those precise primitive types yet, so 
> I'll let you handle that.

No problem, I'll fix it.

Vincent